Agent compatibility: audit, CLI ergonomics, Ollama fixes, cross-tool portability, and test coverage

.agents/skills

@@ -0,0 +1 @@
+../plugins/committy/skills

.claude/hooks/block-dangerous-git.sh

@@ -1,25 +1,61 @@
 #!/bin/bash
+# Block genuinely destructive git operations.
+#
+# Design notes:
+#   - Matches the *command*, not the raw tool input. The previous version
+#     grepped the whole string, so prose in a heredoc that merely named a
+#     command was blocked (including this file's own replacement), while
+#     `git  push` with two spaces or `git -C . push` slipped straight through.
+#   - Plain `git push` is not destructive and is no longer blocked. The
+#     repository's own pre-push hook is the policy-backed control; this hook
+#     only refuses operations that destroy history or uncommitted work.
+#   - Degrades safely when jq is unavailable.
 
 INPUT=$(cat)
-COMMAND=$(echo "$INPUT" | jq -r '.tool_input.command')
-
-DANGEROUS_PATTERNS=(
-  "git push"
-  "git reset --hard"
-  "git clean -fd"
-  "git clean -f"
-  "git branch -D"
-  "git checkout \."
-  "git restore \."
-  "push --force"
-  "reset --hard"
-)
-
-for pattern in "${DANGEROUS_PATTERNS[@]}"; do
-  if echo "$COMMAND" | grep -qE "$pattern"; then
-    echo "BLOCKED: '$COMMAND' matches dangerous pattern '$pattern'. The user has prevented you from doing this." >&2
-    exit 2
-  fi
-done
+
+if command -v jq >/dev/null 2>&1; then
+  COMMAND=$(printf '%s' "$INPUT" | jq -r '.tool_input.command // empty')
+else
+  COMMAND=$(printf '%s' "$INPUT" | sed -n 's/.*"command"[[:space:]]*:[[:space:]]*"\(.*\)".*/\1/p')
+fi
+
+[ -z "$COMMAND" ] && exit 0
+
+# Drop heredoc bodies, so documentation or test data that names a command is
+# not mistaken for an invocation of it. A heredoc body spans lines, so this
+# keeps the line that opens it and discards everything after.
+#
+# Exception: if the heredoc feeds an interpreter, the body IS executed, and
+# dropping it would be a bypass (`bash <<EOF` / `git reset --hard` / `EOF`).
+# In that case scan the whole thing.
+HEREDOC_LINE=$(printf '%s' "$COMMAND" | grep -m1 '<<')
+if [ -n "$HEREDOC_LINE" ] && printf '%s' "$HEREDOC_LINE" |
+  grep -qE '(^|[;&|[:space:]])(bash|sh|zsh|dash|ksh|python3?|perl|ruby|eval|xargs|source)([[:space:]]|$)'; then
+  SCRUBBED="$COMMAND"
+else
+  SCRUBBED=$(printf '%s' "$COMMAND" | awk '/<</ { print; exit } { print }')
+fi
+
+# Only inspect commands that actually invoke git.
+printf '%s' "$SCRUBBED" | grep -qE '(^|[;&|[:space:]])git([[:space:]]|$)' || exit 0
+
+deny() {
+  echo "BLOCKED: '$COMMAND' looks like $1. The user has prevented you from doing this." >&2
+  exit 2
+}
+
+# --force rewrites remote history unconditionally; --force-with-lease is guarded.
+# Checked as two independent conditions rather than one ordered regex: the flag
+# may precede or follow the remote, and `push -f` leaves no space for a pattern
+# that expects one before the flag.
+if printf '%s' "$SCRUBBED" | grep -qE '(^|[;&|[:space:]])push([[:space:]]|$)' &&
+  printf '%s' "$SCRUBBED" | grep -qE '(^|[[:space:]])(--force|-f)([[:space:]]|$)'; then
+  printf '%s' "$SCRUBBED" | grep -q -- '--force-with-lease' || deny "an unguarded force-push"
+fi
+
+printf '%s' "$SCRUBBED" | grep -qE 'reset[[:space:]]+--hard' && deny "a hard reset (discards working-tree changes)"
+printf '%s' "$SCRUBBED" | grep -qE 'clean[[:space:]]+-[a-zA-Z]*f' && deny "a forced clean (deletes untracked files)"
+printf '%s' "$SCRUBBED" | grep -qE 'branch[[:space:]]+-D([[:space:]]|$)' && deny "a forced branch delete"
+printf '%s' "$SCRUBBED" | grep -qE '(checkout|restore)[[:space:]]+(--[[:space:]]+)?\.([[:space:]]|$)' && deny "a bulk discard of working-tree changes"
 
 exit 0

.github/copilot-instructions.md

@@ -0,0 +1,6 @@
+# Copilot instructions
+
+See [`AGENTS.md`](../AGENTS.md) at the repository root. It is the canonical
+guide for every coding agent working in this repository and applies in full:
+build and verify commands, the CLI-first agent workflow, architecture, and the
+commit and pull-request rules.

.github/workflows/cargo-release.yml

@@ -20,7 +20,7 @@ jobs:
       draft: ${{ steps.release_type.outputs.draft }}
       changelog: ${{ steps.changelog.outputs.changelog }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
         with:
           fetch-depth: 0
 
@@ -84,7 +84,7 @@ jobs:
             asset_suffix: macos-arm64
 
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
 
       - name: Get binary name and version
         shell: bash

.github/workflows/lint.yml

@@ -4,7 +4,11 @@ on:
   push:
     branches: [main, develop]
   pull_request:
-    branches: ["*"]
+    # `**` rather than `*`: in Actions branch filters `*` does not match `/`,
+    # so a pull request targeting any branch with a slash in its name -- every
+    # stacked pull request, and anything under a `feat/` or `fix/` prefix --
+    # ran no checks at all.
+    branches: ["**"]
 
 env:
   CARGO_TERM_COLOR: always
@@ -13,7 +17,7 @@ jobs:
   lint-and-test:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
       - name: Install Rust
         uses: actions-rs/toolchain@v1.0.7
         with:

.github/workflows/native-git-verification.yml

@@ -7,6 +7,6 @@ jobs:
   native-git-e2e:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
       - name: Run native git Docker verification
         run: ./scripts/verify_native_git_e2e.sh

AGENTS.md

@@ -6,6 +6,7 @@ Committy is a Rust CLI for conventional commits, branch policy, grouped commits,
 
 - Discover active policy with `cargo run -- schema --output json`.
 - Prefer `--non-interactive`, `--repo-path`, `--dry-run`, and `--output json`.
+- Set `COMMITTY_NONINTERACTIVE=1` in the environment instead of relying on flag placement. `--non-interactive` and `-q`/`--quiet` are global and work on either side of the subcommand; every other flag (`--repo-path`, `--output`, `--dry-run`, and `--verbose`) must follow it.
 - Inspect `api_version`, `ok`, `errors`, and the command-specific plan before applying it.
 - Use the `committy-branch`, `committy-commit`, `committy-release`, and `committy-enforce` plugin skills when available.
 - Never infer permission for a push, publish, destructive Git action, or unrelated staging.
@@ -30,4 +31,4 @@ Tests are in `tests/`; CLI integration tests use `assert_cmd` and temporary repo
 - `src/workflow/`, `src/versioning/`, `src/packages/`: release side effects
 - `plugins/committy/`: shared Codex and Claude skills
 
-PRs target `develop`. Commits follow Conventional Commits. The canonical agent reference is `docs/src/content/docs/reference/agent-workflows.mdx`.
+PRs target `develop`. Commits follow Conventional Commits. Never add AI attribution trailers — no `Co-Authored-By:`, no `Claude-Session:`, no equivalent — to commit messages or PR descriptions, even when tooling suggests them. The canonical agent reference is `docs/src/content/docs/reference/agent-workflows.mdx`.

CLAUDE.md

@@ -1,33 +1,7 @@
-# Committy Claude guide
+@AGENTS.md
 
-Committy is a Rust CLI for conventional commits, branch policy, grouped commits, releases, and multi-package version updates. The checkout is CLI-first; do not assume a TUI exists.
+## Claude Code specifics
 
-## Work safely
+`AGENTS.md` above is the canonical guide and applies in full. Only this section is Claude-specific.
 
-- Discover active policy with `cargo run -- schema --output json`.
-- Prefer `--non-interactive`, `--repo-path`, `--dry-run`, and `--output json`.
-- Inspect `api_version`, `ok`, `errors`, and the command-specific plan before applying it.
-- Use `/committy:committy-branch`, `/committy:committy-commit`, `/committy:committy-release`, and `/committy:committy-enforce` when the plugin is installed.
-- Never infer permission for a push, publish, destructive Git action, or unrelated staging.
-
-## Build and verify
-
-```bash
-cargo check
-cargo test
-cargo fmt --all -- --check
-cargo clippy --all-targets --all-features
-```
-
-Tests are in `tests/`; CLI integration tests use `assert_cmd` and temporary repositories. Add public-seam tests for machine output, exit codes, Git behavior, hook protocol, and plugin manifests.
-
-## Architecture
-
-- `src/cli/commands/`: command implementations and JSON payloads
-- `src/config/`, `src/convention/`: hierarchical typed policy
-- `src/git/`: repository and Git operations
-- `src/linter/`: message/history validation; exit code 3 means lint findings
-- `src/workflow/`, `src/versioning/`, `src/packages/`: release side effects
-- `plugins/committy/`: shared Codex and Claude skills
-
-PRs target `develop`. Commits follow Conventional Commits. The canonical agent reference is `docs/src/content/docs/reference/agent-workflows.mdx`.
+- Invoke the bundled skills as `/committy:committy-branch`, `/committy:committy-commit`, `/committy:committy-release`, and `/committy:committy-enforce` when the plugin is installed. Other runtimes resolve them by bare name.

GEMINI.md

@@ -0,0 +1,5 @@
+# Gemini instructions
+
+See [`AGENTS.md`](./AGENTS.md). It is the canonical guide for every coding agent
+working in this repository and applies in full: build and verify commands, the
+CLI-first agent workflow, architecture, and the commit and pull-request rules.